CN103701475B - Decoding method for Turbo codes with word length of eight bits in mobile communication system - Google Patents

Abstract

A method for realizing Turbo decoding of 8-bit operation word length in a mobile communication system, comprising the following steps: (1) initializing and setting three input data entering into a Turbo decoder, so that the average amplitude thereof is adjusted to that of Turbo decoding Within the safe range, avoid increasing the possibility of overflow due to too large or too small data. (2) Execute the 8bit-RSC recursive systematic convolution coding RSC decoding algorithm with an overflow protection mechanism to obtain the decoding output E ₁ . (3) According to the agreement, execute the interleaving algorithm on the decoding output E ₁ and the array s respectively. (4) Execute the 8bit‑RSC recursive systematic convolution coding RSC decoding algorithm with an overflow protection mechanism to obtain the decoding output E ₂ . (5) Execute the de-interleaving algorithm, judge whether to terminate the iterative operation, and obtain the final decoding result. The present invention is an improved method on the basis of traditional turbo decoding, greatly reduces the overhead of turbo decoding data storage space, and improves the operating efficiency and speed of decoding without obvious loss of decoding performance.

Description

Decoding Method of 8-bit Operation Word Length Turbo Code in Mobile Communication System

technical field

The invention relates to a decoding method of an 8-bit operation word length Turbo code in a mobile communication system, and belongs to the technical field of wireless mobile communication.

Background technique

Turbo (topological) codes were first presented at the ICC International Conference in 1993, by two professors Glavieux A and Thitimajshiwa P of the British University of Communications in France and their Burmese doctoral student Berrou C in the paper "Near Shannonlimit error-correcting coding and decoding :Turbo-codes" was first proposed. The simulation results show that when using a random interleaver with a capacity of 65536 and performing 18 decoding iterations and the signal-to-noise ratio is less than 0.7dB, the bit error rate of the Turbo code with a code rate of 1/2 in the AWGN channel is less than 10 ^{- 5} , close to the Shannon limit (the Shannon limit of 1/2 code rate is 0dB). Due to its excellent performance, Turbo code has become an important coding method for high-speed data services in the third generation and subsequent mobile communication systems.

The decoding of Turbo codes is more complex than its encoding, and its core part is two "soft input soft output (SISO)" decoders, which can significantly improve the bit error performance through multiple iterative decoding. The specific bit error algorithms include: maximum a posteriori probability algorithm MAP and its improved algorithm Log-MAP algorithm and so on. Due to the need for a large number of multiplication and exponent operations, the MAP algorithm is not suitable for hardware implementation. Therefore, some scholars proposed a simplified version of the MAP algorithm in the logarithmic domain: the Log-MAP algorithm. It avoids exponent operations by converting to logarithmic domain operations, and at the same time converts multiplication into addition, and addition becomes Max, that is, the maximum value operation.

On a fixed-point digital signal processor, it is extremely challenging to implement turbo code decoding. In addition to the high complexity of the algorithm itself, another important reason is that the data values are continuously accumulated during the decoding iterative calculation process, which increases the risk of fixed-point number overflow. , which significantly deteriorates the decoding performance.

In order to prevent overflow of fixed-point numbers in the iterative process of the algorithm, Turbo decoding usually chooses an operation word length of more than 8 bits. Taking the 16-bit word length as an example, the value range of the data is between -32768 and 32767, while the data value of the 8-bit word length is between -128 and 127, and the range is extremely small. Therefore, the main problem of Turbo decoding with 8-bit operation word length is: as the iterative operation continues to deepen, the data value will continue to increase, and overflow is more likely to occur.

However, using an 8-bit operation word length can reduce the data storage space of the Turbo decoding algorithm. At the same time, the algorithm will be more conducive to parallel processing, so that the algorithm execution efficiency and speed will be much better than the decoding algorithm with an operation word length of 8 bits or more. Especially as the mobile communication system has higher and higher requirements on the rate and efficiency of real-time signal processing, Turbo decoding with 8-bit operation word length is very attractive.

Referring to Fig. 1, the implication of each character in this figure is introduced: I _k is the systematic bit in the data to be decoded that obtains from the channel at the k moment, and the natural number k subscript is the time sequence number, is the first parity bit in the data to be decoded obtained from the channel at the kth moment, is the second parity bit in the data to be decoded obtained from the channel at the kth moment, R _k is the data to be decoded obtained from the channel at the kth moment, and the data to be decoded includes the above three kinds of data: I _k , and E ₁ is the external information data output by the first decoder, is the external information data output by the first decoder and the interleaved data, E ₂ is the external information data output by the second decoder, It is the external information data output by the second decoder and the deinterleaved data.

Referring to Fig. 2, introduce a kind of typical Turbo code decoder structure and its decoding method shown in Fig. 2 again: at first with the received data y _k to be decoded: promptly the information bit I _k of input, the first verification bits together into the first decoder. The interleaved data and the second parity bits of I _k together into the second decoder. The extrinsic data E ₁ output by the first decoder is also sent to the second decoder for decoding after interleaving, and the extrinsic data E ₂ output by the second decoder is sent to the first decoder after deinterleaving device to decode. Turbo code decoding is an iterative process, and the usual number of iterations is 3 to 8 times. In the last iteration, the second decoder will output the log likelihood ratio LLR (Loglikelihood Ratio), and after deinterleaving, perform The hard decision is used as the final decoding result. During the decoding process, the value of each variable is a signed integer, and the value in the 16-bit word-length decoding operation is between -32768 and 32767, and the value in the 8-bit word-length decoding operation is - Between 128 and 127.

The final result of decoding is to calculate the logarithmic likelihood ratio LLR, the calculation formula of the logarithmic likelihood ratio is: In the formula, α _k is the forward metric, β _k is the backward metric, and γ is the transition probability. In this way, as long as α _k , β _k and γ are calculated, the LLR can be calculated. Moreover, the method of calculating α _k , β _k , and γ can be calculated by using the traditional maximum a posteriori probability MAP calculation method, which will not be repeated here.

In a word, Turbo code decoding is a complex operation process, not only the algorithm operation itself is very complicated, but also due to multiple iterations and recursion, the forward metric α _k and the backward metric β _k in the calculation process are continuously accumulated and increased , which poses a great risk of overflow for a given number of points.

In order to reduce the storage space overhead of the Turbo decoding algorithm and improve its execution efficiency and speed, an effective method is to choose a smaller operation word length. However, the ensuing problem is: the smaller the operation word length is, the more likely the overflow of the operation word length will occur as the decoding algorithm iteratively deepens and the data value continues to increase. At present, the turbo decoding of 8-bit operation word length is known as the minimum word length that can be practically applied, and its data value range is between -128 and 127, which is extremely small, so overflow is very prone to occur.

But so far, no perfect solution has been found for the overflow problem of Turbo decoding with 8-bit operation word length, which is also the biggest difficulty in realizing Turbo decoding with 8-bit operation word length.

Contents of the invention

In view of this, the purpose of the present invention is to provide a method for realizing Turbo decoding of 8-bit operation word length in a mobile communication system. The present invention is an improved method based on traditional Turbo decoding, and realizes 8 bits well. The turbo decoding of the word length of the bit operation greatly reduces the overhead of the data storage space of the turbo decoding without obvious loss of decoding performance, and improves the operation efficiency and speed of the decoding.

In order to achieve the above object, the present invention provides a method for realizing Turbo decoding of 8-bit operation word length in a mobile communication system. The method is realized based on the traditional Turbo decoding MAX-Log-MAP forward and backward metric calculation method. Decoding has three input data that are 1×K dimensional arrays: systematic bit Rs, first parity bit Rp ₁ and second parity bit Rp ₂ , where K is the total number of information bits, that is, the decoded output data Length, 3 arrays Rs, Rp ₁ , the element of Rp ₂ is all 8 bit word lengths; It is characterized in that: described method comprises the following operation steps:

Step 1. Initialize the data to be entered into the Turbo decoder, so that the average amplitudes of the three input data are all adjusted within the safe range of Turbo decoding, avoiding the possibility of overflow due to too large or too small data property; this step 1 includes the following operations:

(11) Define three 1×K-dimensional arrays s, p ₁ , p ₂ , and initialize and set the three arrays s=Rs, p ₁ =Rp ₁ and p ₂ =Rp ₂ respectively; among them, the array s is used for temporary Storage system bits, p ₁ is used to temporarily store the first parity bit, and p ₂ is used to temporarily store the second parity bit;

(12) Define five 1×K-dimensional arrays Δs, E ₁ , ΔE ₁ , E ₂ , ΔE ₂ , and set the initial values of all elements of the five arrays to zero; where Δs is the interleaved value of s Array, E ₁ is the output of the recursive system convolution coding RSC decoding algorithm when the input is s, p ₁ , ΔE ₂ , ΔE ₁ is the array after E ₁ is interleaved, E ₂ is when the input is s, p ₂ , ΔE ₁ The output of the RSC decoding algorithm, ΔE ₂ is the array after E ₂ deinterleaving;

(13) Set the element word length of eight arrays s, Δs, p ₁ , p ₂ , E ₁ , ΔE ₁ , E ₂ , ΔE ₂ to 8 bits;

(14) Calculate the average amplitude of all elements in the three arrays s, p ₁ , p ₂ after the initial setting in step (11), the calculation method of the average amplitude is: first calculate the three arrays s, p ₁ , The magnitude sum of all elements in each array in p ₂ , and then divide the magnitude sum by the number of elements in the array to obtain the quotient A; then, according to the following formula: Scale the amplitude of the three arrays s, p ₁ and p ₂ respectively, so that the average amplitude of all elements in s, p ₁ and p ₂ becomes 32;

Step 2, execute the recursive system convolution coding RSC (RecursiveSystematic Convolutional code) decoding algorithm of 8bit-RSC with overflow protection mechanism, obtain the decoding output E ₁ ; This step 2 includes the following operations:

(21) Judge whether the absolute value in the element of the array ΔE ₂ exceeds the threshold value N, and the numerical value of the threshold value can be adjusted flexibly; if so, perform a divisive normalization operation before decoding: each array s, All elements in p ₁ , p ₂ , ΔE ₂ are divided by the normalization coefficient μ, μ>1; if not, no processing is done;

(22) With s, p ₁ , ΔE ₂ as input parameters, execute the 8bit-RSC decoding algorithm based on the traditional MAX-Log-MAP forward and backward measurement calculation method and improve, and obtain the decoding output E ₁ ; The key operation of the improved 8bit-RSC decoding algorithm is the calculation of the forward and backward metrics. The improvements are: the internal data stored in the calculation process are all set to 8-bit word length, and the subtraction return is added in the decoding process. One operation;

(22) With s, p ₁ and ΔE ₂ as input parameters, perform an improved 8bit-RSC decoding algorithm to obtain a decoding output E ₁ ;

Step 3, according to the 3GPP protocol, perform an interleaving algorithm on the decoding output E1 and the array _s respectively; this step 3 includes the following operations:

(31) using the decoding output E1 of step ( ₂ ) as the interleaving input, calling the interleaving algorithm according to the 3GPP agreement, and assigning the interleaving output to ΔE1 _;

(32) Using the array s as the interleaving input, after performing the same interleaving algorithm as step (31) again, assign the interleaving output to Δs;

Step 4, execute the recursive systematic convolutional coding RSC decoding algorithm of 8bit-RSC with an overflow protection mechanism again, and obtain the decoding output E ₂ ; this step 4 includes the following operations:

(41) Determine whether the absolute value of the elements in the array ΔE ₁ exceeds the threshold value N, if so, perform a divisive normalization operation before decoding: four arrays s, p ₁ , p ₂ , ΔE ₁ All elements of are divided by the normalization coefficient μ;

(42) Take s, p ₂ , ΔE ₁ as input, implement the traditional MAX-Log-MAP forward and backward metric calculation method, and make the following improvements: make new conventions and introduce new functions to the RSC decoding algorithm, and get 8bit-RSC decoding algorithm, the new conventions and the introduction of new functions include: the internal data stored in the calculation process are all 8-bit word length, and the subtraction normalization operation is added in the decoding process:

The specific calculation method is: first set the internal data storage as 8-bit word length, then calculate the forward metric value and the backward metric value respectively according to the description of step (22), and then according to the obtained forward and backward metric values and the traditional MAX- The Log-MAP forward and backward metric calculation method calculates the decoding output value and assigns it to _E2 ;

Step 5, execute the deinterleaving algorithm, judge whether to terminate the iterative operation, and obtain the final decoding result; this step 5 includes the following operations:

(51) With the output E ₂ of step 4 as input, call the de-interleaving algorithm stipulated in the 3GPP protocol, perform the de-interleaving operation, and assign the output to ΔE ₂ ;

(52) According to the traditional Turbo decoding termination strategy, it is judged whether the iteration termination condition has been met; if the condition is satisfied, the final decoding result is obtained according to _E2 calculation, and the entire Turbo decoding algorithm is ended; otherwise, return to step 2 and continue again implement.

The implementation method of 8-bit operation word length Turbo decoding in the mobile communication system of the present invention proposes a new comprehensive scheme for solving the Turbo iterative fixed-point overflow problem, and its key innovative technology has three main points:

1. The operation word length of Turbo decoding is compressed to 8 bits, which not only reduces data storage overhead, but also facilitates parallel processing.

2. Adjust the average magnitude of the input data.

3. On the basis of the traditional Turbo decoding MAX-Log-MAP forward and backward metric calculation method, an 8bit-RSC decoding algorithm with an overflow protection mechanism is proposed.

The advantage of the method for realizing Turbo decoding with 8-bit operation word length in the mobile communication system of the present invention is: compared with Turbo decoding with more than 8-bit operation word length, the method of the present invention significantly reduces data storage overhead, improves operation speed and efficiency. For the overflow problem common to Turbo, the overflow prevention effect of the method of the present invention is obvious.

Description of drawings

FIG. 1 is a schematic flowchart of a traditional turbo decoding implementation method.

Fig. 2 is a schematic flow chart of the improved method of turbo decoding with 8-bit operation word length according to the present invention.

Fig. 3 is a flow chart of specific operation steps of the Turbo decoding method for 8-bit operation word length according to the present invention.

Fig. 4 is a schematic diagram of adding a subtraction normalization operation on the basis of the traditional MAX-Log-MAP forward metric calculation method in the 8-bit operation word length Turbo decoding method of the present invention to reduce the possibility of data overflow.

Fig. 5 is a schematic diagram of adding a subtraction normalization operation on the basis of the traditional MAX-Log-MAP backward metric calculation method in the 8-bit operation word length Turbo decoding method of the present invention to reduce the possibility of data overflow.

Fig. 6 is a schematic diagram of the mathematical meaning of the subtraction module in Fig. 3 and Fig. 4 .

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

Turbo decoding has three input data, namely: systematic bit Rs, first parity bit Rp ₁ and second parity bit Rp ₂ , they are all 1×K dimensional arrays; where K is the total number of information bits, also It is the data length of the decoded output, and the elements of the three arrays Rs, Rp ₁ , and Rp ₂ are all 8-bit words long.

Referring to Fig. 2 and Fig. 3, introduce the implementation method of 8 bit operation word length Turbo decoding in the mobile communication system of the present invention, it is to improve and realize on the basis of traditional Turbo decoding MAX-Log-MAP forward and backward metric calculation method of; the method includes the following steps:

Step 1. Initialize the data to be entered into the Turbo decoder, so that the average amplitudes of the three input data are all adjusted within the safe range of Turbo decoding, avoiding the possibility of overflow due to too large or too small data sex. This step 1 includes the following operations:

(11) Define three 1×K-dimensional arrays s, p ₁ , p ₂ , and initialize and set the three arrays s=Rs, p ₁ =Rp ₁ and p ₂ =Rp ₂ respectively; among them, the array s is used for temporary Storage system bits, p ₁ is used to temporarily store the first parity bit, and p ₂ is used to temporarily store the second parity bit.

(12) Define five 1×K-dimensional arrays Δs, E ₁ , ΔE ₁ , E ₂ , ΔE ₂ , and set the initial values of all elements of the five arrays to zero; where Δs is the interleaved value of s Array, E ₁ is the output of the recursive system convolution coding RSC decoding algorithm when the input is s, p ₁ , ΔE ₂ , ΔE ₁ is the array after E ₁ is interleaved, E ₂ is when the input is s, p ₂ , ΔE ₁ The output of the RSC decoding algorithm, ΔE ₂ is the array after E ₂ deinterleaving.

(13) Set the element word lengths of the eight arrays s, Δs, p ₁ , p ₂ , E ₁ , ΔE ₁ , E ₂ , ΔE ₂ to 8 bits.

(14) Calculate the average amplitude of all elements in the three arrays s, p ₁ and p ₂ after step (11) initialization, the calculation method of the average amplitude is: first calculate the amplitude and sum of the three array elements respectively , and then divided by the number of array elements; then, according to the following formula: Perform magnitude scaling on the three arrays s, p ₁ , p ₂ respectively, so that the average magnitude of all elements in s, p ₁ , p ₂ becomes 32.

Step 2, execute the 8bit-RSC recursive systematic convolution coding RSC decoding algorithm with an overflow protection mechanism, and obtain the decoding output E ₁ . This step 2 includes the following operations:

(21) Judging whether the absolute value of the elements of the array ΔE ₂ exceeds the threshold value N, the value of the threshold value can be adjusted flexibly, and the threshold value N is usually set to 96; if so, perform divisive regression before decoding Normalization operation: Divide all elements in each array s, p ₁ , p ₂ , ΔE ₂ by the normalization coefficient μ, μ>1, set to 2. If not, do nothing.

(22) With s, p ₁ and ΔE ₂ as input parameters, execute the 8bit-RSC decoding algorithm with overflow protection mechanism to obtain the decoding output E ₁ .

In the 8bit-RSC decoding algorithm, the most critical calculation step is the calculation of forward and backward metrics. Therefore this step (22) is the key operation in the method of the present invention, and its improvement is: on the basis of the traditional MAX-Log-MAP forward and backward measurement calculation method, new agreement and change have been made: the one, in the calculation process The stored internal data are all set to 8-bit word length, and the second is to add a subtraction normalization operation in the decoding process. (See Figure 3, Figure 4, and Figure 5 for how and where the new subtractive normalization works).

This step (22) includes the following specific operations:

(22a) First set the internal data stored in the calculation process as 8-bit word length;

(22b) Use the traditional MAX-Log-MAP forward metric calculation method to calculate the forward metric value: there are 8 forward metric values corresponding to the decoding of the i-th information bit, and they are represented by the array α[i][m] Among them, α represents an array, and the natural numbers i and m are the sequence numbers of information bits and forward metric values respectively, and their maximum values are K and 8 respectively; is a temporary array; the array γ[i][n] is the 4 branch metrics of the i-th information bit, where the natural numbers i and n are the sequence numbers of another information bit and another forward metric value, and the largest The values are K and 4 respectively;

The calculation method of the improved MAX-Log-MAP forward metric is: according to the traditional MAX-Log-MAP backward metric calculation method, first calculate α[i-1][m] and γ[i-1][n] Perform accumulation, comparison and selection operations, and calculate Then, use the formula Perform subtraction and normalization operations to obtain α[i][m];

(22c) Use the improved MAX-Log-MAP backward metric calculation method to calculate the backward metric value: there are 8 backward metric values corresponding to the jth information bit after decoding, and they are respectively represented by the array β[j][q] Among them, β represents an array, and the natural numbers j and q are the sequence numbers of information bits and backward metric values respectively, and their maximum values are K and 8 respectively; is a temporary array;

The improved MAX-Log-MAP backward metric calculation method is: first perform the operations of accumulating, comparing and selecting β[j][q] and γ[j][n] according to the addition operation, and then according to the traditional MAX-Log -MAP calculation method is calculated from β[j][q] Then, use the formula Perform subtraction and normalization operations to obtain β[j-1][q];

(22d) Calculate the decoding output according to the obtained forward and backward metrics according to the traditional method, and assign it to E ₁ .

Step 3, according to the 3GPP protocol, perform an interleaving algorithm on the decoded output E ₁ and the array s respectively. This step 3 includes the following operations:

(31) Using the decoding output E ₁ of step (2) as an interleaving input, calling an interleaving algorithm according to the 3GPP protocol, and assigning the interleaving output to ΔE ₁ .

(32) Using the array s as the interleaving input, after performing the same interleaving algorithm as in step (31) again, assign the interleaving output to Δs.

Step 4, execute the 8bit-RSC recursive systematic convolution coding RSC decoding algorithm with an overflow protection mechanism again, and obtain the decoding output E ₂ . This step 4 includes the following operations:

(41) Determine whether the absolute value of the elements in the array ΔE ₁ exceeds the threshold value N, and if so, perform a divisive normalization operation before decoding: combine the four arrays s, p ₁ , p ₂ , ΔE ₁ All elements in are divided by the normalization coefficient μ; where, the threshold value N is 96, and the normalization coefficient μ must be greater than 1, which is set to 2.

(42) With s, p ₂ , ΔE ₁ as input, execute the traditional MAX-Log-MAP forward and backward metric calculation method, and make the following improvements: 8bit-RSC decoding algorithm that makes new conventions and introduces new functions: The internal data stored in the calculation process are all 8-bit word length, and the subtraction normalization operation is added in the decoding process:

The specific calculation method is as follows: first set the internal data storage to be 8-bit word length, then calculate the forward metric value and the backward metric value respectively according to the description of step (22), and then calculate according to the obtained forward and backward metric values and the traditional method Decode the output value and assign it to E ₂ .

Step 5, execute the de-interleaving algorithm, judge whether to terminate the iterative operation, and obtain the final decoding result. This step 5 includes the following operations:

(51) With the output E ₂ of step 4 as input, call the de-interleaving algorithm stipulated in the 3GPP protocol, perform the de-interleaving operation, and assign the output to ΔE ₂ ;

(52) According to the traditional Turbo decoding termination strategy, it is judged whether the iteration termination condition has been met; if the condition is satisfied, the final decoding result is obtained according to _E2 calculation, and the entire Turbo decoding algorithm is ended; otherwise, return to step 2 and continue again implement.

The present invention has carried out the simulation test of embodiment for many times, and the result of test is successful, has realized the purpose of the invention.

Claims (3)

1. An implementation method of 8-bit operation word length Turbo decoding in a mobile communication system, which is based on the traditional Turbo decoding MAX-Log-MAP forward and backward metric calculation method; Turbo decoding has three 1 × K Input data of two-dimensional array: systematic bit Rs, first parity bit Rp ₁ and second parity bit Rp ₂ , where K is the total number of information bits, that is, the data length of decoding output, three arrays Rs, Rp ₁ , the elements of Rp ₂ are all 8-bit word lengths; it is characterized in that: the method comprises the following steps: Step 1. Initialize the data to be entered into the Turbo decoder, so that the average amplitudes of the three input data are all adjusted within the safe range of Turbo decoding, avoiding the possibility of overflow due to too large or too small data property; this step 1 includes the following operations: (11) Define three 1×K-dimensional arrays s, p ₁ , p ₂ , and initialize and set the three arrays s=Rs, p ₁ =Rp ₁ and p ₂ =Rp ₂ respectively; among them, the array s is used for temporary Storage system bits, p ₁ is used to temporarily store the first parity bit, and p ₂ is used to temporarily store the second parity bit; (12) Define five 1×K-dimensional arrays Δs, E ₁ , ΔE ₁ , E ₂ , ΔE ₂ , and set the initial values of all elements of the five arrays to zero; where Δs is the interleaved value of s Array, E ₁ is the output of the recursive system convolution coding RSC decoding algorithm when the input is s, p ₁ , ΔE ₂ , ΔE ₁ is the array after E ₁ is interleaved, E ₂ is when the input is s, p ₂ , ΔE ₁ The output of the RSC decoding algorithm, ΔE ₂ is the array after E ₂ deinterleaving; (13) Set the element word length of eight arrays s, Δs, p ₁ , p ₂ , E ₁ , ΔE ₁ , E ₂ , ΔE ₂ to 8 bits; (14) Calculate the average amplitude of all elements in the three arrays s, p ₁ , p ₂ after the initial setting in step (11), the calculation method of the average amplitude is: first calculate the three arrays s, p ₁ , The magnitude sum of all elements in each array in p ₂ , and then divide the magnitude sum by the number of elements in the array to obtain the quotient A; then, according to the following formula: Scale the amplitude of the three arrays s, p ₁ and p ₂ respectively, so that the average amplitude of all elements in s, p ₁ and p ₂ becomes 32; Step 2, execute the recursive system convolution coding RSC (RecursiveSystematic Convolutional code) decoding algorithm of 8bit-RSC with overflow protection mechanism, obtain the decoding output E ₁ ; This step 2 includes the following operations: (21) Judge whether the absolute value in the element of the array ΔE ₂ exceeds the threshold value N, and the numerical value of the threshold value can be adjusted flexibly; if so, perform a divisive normalization operation before decoding: each array s, All elements in p ₁ , p ₂ , ΔE ₂ are divided by the normalization coefficient μ, μ>1; if not, no processing is done; (22) With s, p ₁ , ΔE ₂ as input parameters, execute the 8bit-RSC decoding algorithm with an overflow protection mechanism, and obtain the decoding output E ₁ ; Step 3, according to the 3GPP protocol, perform an interleaving algorithm on the decoding output E1 and the array _s respectively; this step 3 includes the following operations: (31) using the decoding output E1 of step ( ₂ ) as the interleaving input, calling the interleaving algorithm according to the 3GPP agreement, and assigning the interleaving output to ΔE1 _; (32) Using the array s as the interleaving input, after performing the same interleaving algorithm as step (31) again, assign the interleaving output to Δs; Step 4, execute the recursive systematic convolutional coding RSC decoding algorithm of 8bit-RSC with an overflow protection mechanism again, and obtain the decoding output E ₂ ; this step 4 includes the following operations: (41) Determine whether the absolute value of the elements in the array ΔE ₁ exceeds the threshold value N, if so, perform a divisive normalization operation before decoding: four arrays s, p ₁ , p ₂ , ΔE ₁ All elements of are divided by the normalization coefficient μ; (42) Take s, p ₂ , ΔE ₁ as input, execute the traditional MAX-Log-MAP forward and backward metric calculation method, and make the following improvements: make new conventions and introduce new functions to the decoding algorithm, and get 8bit- RSC decoding algorithm, the new conventions and the introduction of new functions include: the internal data stored in the calculation process are all 8-bit word length, and the subtraction normalization operation is added in the decoding process: The specific calculation method is: first set the internal data storage as 8-bit word length, then calculate the forward metric value and the backward metric value respectively according to the description of step (22), and then according to the obtained forward and backward metric values and the traditional MAX- The Log-MAP forward and backward metric calculation method calculates the decoding output value and assigns it to _E2 ; Step 5, execute the deinterleaving algorithm, judge whether to terminate the iterative operation, and obtain the final decoding result; this step 5 includes the following operations: (51) With the output E ₂ of step 4 as input, call the de-interleaving algorithm stipulated in the 3GPP protocol, perform the de-interleaving operation, and assign the output to ΔE ₂ ; (52) According to the traditional Turbo decoding termination strategy, it is judged whether the iteration termination condition has been met; if the condition is satisfied, the final decoding result is obtained according to _E2 calculation, and the entire Turbo decoding algorithm is ended; otherwise, return to step 2 and continue again implement. 2. The method according to claim 1, characterized in that: the key operation in the 8bit-RSC decoding algorithm with overflow protection mechanism is the calculation of forward and backward metrics, and its improvement is: stored in the calculation process Internal data is all set to 8 bit word length, and new subtraction normalization operation is added in the decoding process; This step (22) comprises following specific operation content: (22a) First set the internal data stored in the calculation process as 8-bit word length; (22b) Use the improved MAX-Log-MAP forward metric calculation method to calculate the forward metric value: there are 8 forward metric values corresponding to the decoding of the i-th information bit, and they are represented by the array α[i][m] Among them, α represents an array, and the natural numbers i and m are the sequence numbers of information bits and forward metric values respectively, and their maximum values are K and 8 respectively; is a temporary array; the array γ[i][n] is the 4 branch metrics of the i-th information bit, where the natural numbers i and n are the sequence numbers of another information bit and another forward metric value, and the largest The values are K and 4 respectively; The calculation method of the improved MAX-Log-MAP forward metric is: according to the traditional MAX-Log-MAP forward metric calculation method, first calculate α[i-1][m] and γ[i-1][n] Perform accumulation, comparison and selection operations, and calculate Then, use the formula Perform subtraction and normalization operations to obtain α[i][m]; (22c) Use the improved MAX-Log-MAP backward metric calculation method to calculate the backward metric value: there are 8 backward metric values corresponding to the jth information bit after decoding, and they are respectively represented by the array β[j][q] Among them, β represents an array, and the natural numbers j and q are the sequence numbers of information bits and backward metric values respectively, and their maximum values are K and 8 respectively; is a temporary array; The improved MAX-Log-MAP backward metric calculation method is: first perform the operations of accumulating, comparing and selecting β[j][q] and γ[j][n] according to the addition operation, and then according to the traditional MAX-Log -MAP backward metric calculation method is calculated from β[j][q] Then, use the formula Perform subtraction and normalization operations to obtain β[j-1][q]; (22d) Calculate the decoding output according to the obtained forward and backward metrics according to the traditional method, and assign it to E ₁ . 3. The method according to claim 1, characterized in that: the threshold value N in the step 2 and step 4 is set to 96, and the normalization coefficient μ is set to 2.